Sharing with the best paper winner of CV Summit: how is a good paper refined?

4 month 23 Day CVPR 2022 The paper pre sharing meeting came to a successful conclusion . Different from previous years , This sharing meeting adopts the activity mode of offline and online linkage , While ensuring the sharing effect, more audiences participated in the activity , In total 9 During the 24-hour live broadcast , The accumulated harvest is as high as 1.5 Million online viewers watch , Average viewing time 36 minute , It reflects a good audience participation .

Except under six themes 20 Speakers shared their latest research results and discussed the current situation and future of the field , This time we are lucky to have Ma Yi 、 Tu Zhuowen 、 Dai Yuchao 、 Zhang Xiangyu and Huang Gao These winners of the best papers of the vision summit over the years , With “ good How is the thesis refined ？” A wonderful round table discussion was held for the theme , Shared their thoughts on “ A good paper ” and “ Good research ” The understanding of the .

What's interesting is that , The research of teachers has certain inheritance . From teacher Ma Yi ICCV99 Of 3D Geometric Research , To his new work on low rank expression in high-dimensional space , I was in 10 Many years ago, it promoted the hottest direction in the visual field at that time ; And teacher Dai Yuchao is CVPR12 Award winning work of , It happens to be 3D The perfect combination and summary of geometry and low rank expression . Since then, with the arrival of the era of deep learning , Many classic jobs have emerged in the gold rush era , Including teacher Tu Zhuowen in 2015 Award winning work in HED, He passed the work and the preface DSN The work tries to solve the problem of effective training of deep network . And this problem is also 2016 Year of ResNet Has been almost perfectly solved ,2016 year CVPR Zhang Xiangyu, the winner of the best paper, shared with you ResNet The story behind it .2017 year CVPR The winner of the best thesis, Mr. Huang Gao, is about DenseNet The research work of , It's the following ResNet Then another attempt of convolution network , And another research work of teacher Huang Gao Stochastic Depth It has become a training vision Transformer Standard configuration of .

Playback video address

Round table forum for the winner of the best paper of the visual summit ： How to refine a good thesis

https://www.bilibili.com/video/BV1Xr4y1J7xq?share_source=copy_web